基于机器学习的无线传感网络通信异常入侵检测技术

Communication Anomaly Intrusion Detection Technology for Wireless Sensor Networks Based on Machine Learning

无线传感网络通信因节点资源有限,易受第三方攻击,因此以机器学习为基础,研究了一种无线传感网络通信异常入侵检测技术。将机器学习中的支持向量机应用于无线传感网络中,在初始权重空间构建支持向量机节点定位模型;明确现阶段网络通信状态;创建节点重要性概念,将重要性强的节点设为易被攻击对象;推算异常行为造成的潜在损失,若大于设定临界值,视该节点为异常入侵并进行隔离。仿真分析表明:所提方法的误报率始终低于6%,漏报率最高为4%,网络能量消耗低于2 J,在25次迭代仿真分析过程中异常入侵检测时间的波动范围为2 s~6 s。仿真结果验证了所提技术具备优秀的异常入侵检测精度与效率,能有效降低网络能耗,鲁棒性强。

Wireless sensor network communication is vulnerable to third-party attack because of its limited node resources, so based on machine learning, a wireless sensor network communication anomaly intrusion detection technology is studied. Support vector machine is applied to wireless sensor network, and the node location model of support vector machine is constructed in the initial weight space. The current network communication state is defined, the concept of node importance is established, and the node with strong importance is set as the easy object to be attacked. The potential loss caused by abnormal behavior is calculated. If it is greater than the critical set value, the node is regarded as abnormal intrusion and isolated. The simulation analysis shows that the false positive rate of the method is always less than 6%,the highest miss detection rate is 4%,the network energy consumption is less than 2 J,and the fluctuation range of abnormal intrusion detection time is 2 s~6 s in the process of 25 iterative simulation analysis. The simulation results show that the proposed method has excellent accuracy and efficiency in anomaly intrusion detection, can effectively reduce network energy consumption, and has strong robustness.